Festival installation

[Tim L]

Hi,

I went through the installation process of Festival on my Ubuntu desktop.
Only after the fact I found out that festival is available by default through Ubuntu’s package manager (sudo apt install festival). Perhaps there could be a heads-up in the tutorial to go that route?

Furthermore, if one installs from source, it should be mentioned that only the North American servers actually have the newest version of festival (2.5) – the European downloads are all at 2.4.
2.5 will install with gcc7 (it’s been updated for gcc6, but it works with 7), 2.4 won’t – and gcc6+ will be installed by default on recent Linux distributions. I didn’t realise this, thinking that the newest source would surely be on CSTR’s servers.

[Dan L]

Did you have to install the Edinburgh Speech Tools Library separately? Or did the package manager version include all necessary dependencies?

Thanks.

http://www.festvox.org/docs/manual-2.4.0/festival_6.html#Installation

[Dan L]

I’ll answer my own question – it worked perfectly and seems to install all the required dependencies. I would say it’s definitely worth updating the documentation with this info. It was quick and easy.

[Tim L]

Hey, does everything work for you when going through the exercise?
I thought my manual installation worked, but fixing error after error finally I can’t get the Python wrappers to work. Also, there seems to be some mismatch in how the manual calls the multisyn-build commands vs. how their usage is defined in my installation.
I’m quite frustrated actually and consider dumping it and using the packaged method, but I don’t want to potentially lose my progress 🙂

[Dan L]

I used `sudo apt install festival` and it pretty much worked straight away I believe. There is an issue in that the setup.sh file contains university server paths which we’ll need to swap out for local paths. I haven’t got to that yet which is preventing me from really playing with it. But out of the box, Festival appears to be creating speech from the default voice no bother.

[Dan L]

Just re-reading your last reply a bit more carefully and yes, perhaps I’m about to discover all the issues you’ve had!

[Tim L]

I actually tried with the ubuntu install – it may work if some time is taken (should also install speech-tools then), but as those are all just binaries, to me it wasn’t clear how to reference the right folders in the setup script, so I abandoned that.
In the end I got it to work by copying the right files from the lab machines, and recompiling on my Ubuntu 18.04 server distribution using gcc-4.8. festvox didn’t compile – here I used the version downloaded from festvox.org, and copied missing files over from the ppls version. I used a python3.6 environment through anaconda, with the pre-compiled shared library from the ppls computers.
Last note: I did not take the symlinked directories in festival/lib, as they’re large. But I did need one directory from inside lib/voices-multisyn to run the voice locally, specifically the one containing the voice_localdir_multisyn-rpx script.

Bottom line:
I don’t recommend doing this to anyone, and I encourage Simon to remove the mention that you could perhaps install this locally, as it is definitely not possible to run through the entire exercise without taking some files from the PPLS computers, unless perhaps you are running a specific old distribution of linux/mac.
What !maybe! works, is following this tutorial (in fact I did for part of the exercise in the beginning), but again, not recommended, as the scripts used are different from the ones living on the PPLS computers.

[Dan L]

Thanks for doing all that investigative work! Guess I’ll stick to the lab computers then…

[Author]

Posts